Tags: academia, musings, research

« Previous post: Writing a curriculum vitæ — Next post: Why your academic code needs a software … »

One of the strangest phenomena in academia is the reliance on publishing companies. In this article, I want to outline some of the issues that arise when working with publishers. I shall also endeavour to provide some solutions to improve this collaboration.

Before we start, a brief disclaimer: this article will use an amalgamation of different incidents that involved either myself, my colleagues, or my friends. Names (of the publishing companies) have been withheld because I do not think it fair to use my ‘soapbox’ without giving the other side a chance to respond. Moreover, everything I write here pertains to publishing your research in a journal. Conference publishing—at least in machine learning—is a joy¹. Plus, there are good examples of journals for machine learning papers, foremost of them the Journal of Machine Learning Research. The ‘adversaries’ in this article are rather the ‘big’ publishing companies and their practices. With that out of the way, let us take a look at the state of the art!

The journal publishing process in academia

If you are new to science, at some point, you will probably have to deal with an established publishing company to get your article published. The deal usually works like this:

1. You look for a journal you want to publish in and submit your article to the journal. This already often involves jumping through some hoops. Without knowing the eventual fate of your article, you often already have to abide by certain arbitrary formatting guidelines or completely ‘butcher’ your article for the submission² by shifting content around. However, this can be all accepted and endured because of course, you want something from them, i.e. a published, citable publication!

2. The journal then receives your submission—often through a web interface that was developed with all the UX/UI knowledge of the 1980s³ and has never been updated since—and this is where it gets slightly murky. Another person—typically the editor of the journal—now decides whether to accept the paper for reviewing, or whether to provide you with a desk reject. A desk reject usually cannot be appealed. It just shows you the door and leaves you to try again with another journal (more murkiness here). For early-career researchers like Ph.D. students submitting their first paper, this can be highly discouraging. I see the reason for reducing the workload on reviewers, of course, but I also heard of academic feuds that were carried out on the backs of Ph.D. students and their publications.

3. Assuming your paper ‘survived’ the desk reject, it will now be sent to reviewers or referees. Their job is to review your paper thoroughly, provide feedback, and in general give this whole business a formal veneer. Setting aside problems in the reviewing process—which I discussed in another blog post—this again opens up a portal into a strange dimension: working as a reviewer for a journal is usually a job that is provided for free (same goes for editorial duties). Notice that this is despite the fact that those journals are charging money to readers and universities. ETH Zurich, my current employer, describes their experiences of the negotiations with publishers and mentions an expenditure of 6.4 million EUR (roughly 7.25 million USD) per year for being allowed to access journal articles. That is a lot of money.

   Setting aside the actual numbers here, let me just point out how strange it is that companies are relying on unpaid labour, and this reliance is crucial to their business model. They often do not employ people that are qualified to judge the content that they want to publish! But of course, reviewers and editors get the benefit of exposure—that wonderful currency that is supposed to help your career along! Even stranger: journals often charge hefty sums for accessing your own research articles. To me, it is super weird that research that is often funded by the taxpayer cannot be accessed by the taxpayer.

4. Supposing your article got sufficiently good reviews to be published, the next stage of the process starts. This is where the meddling begins in earnest. After a little back and forth, you article will be changed according to some arbitrary rules: the last period of every sentence in an image caption will be removed, footnotes will be put into the text—because for some reason, footnotes are permitted in virtually every template and publishing medium, but deemed somewhat uncouth by certain publishers—and you might have to redo certain parts of your paper because of subtle font changes or what have you.

   Again, lest you think of me as a particularly cranky person prone to grumbling and finding faults, you are getting the wrong idea here. I do not object to these changes, but I do object to the fact that these meddlesome changes often decrease the quality of your paper. Here are some irksome changes:

   • Footnotes will be inserted willy-nilly into the text, regardless of whether they make sense or not. That might break the flow of your paper, but that is your problem.

   • Some ‘publisher house rules’ conflict with proper nomenclature in a field. For example, the journal might have the ‘rule’ that all fields in a table have to be capitalised. If this clashes with nomenclature in your field, it is—you guessed it—your problem.

   • Your equations will typically be typeset yet another time for you⁴, which might introduce subtle changes: symbols will change and you have to be go through your own paper once again line-by-line to see whether anything untoward happened. Again, I am primarily objecting to the substandard quality of this meddling: the work that you put into writing your equations is completely ignored, and now you have to chase—often very subtle—changes in your own text. For mathematical typesetting, precision is crucial, and it is unbecoming when people who do not care about this precision create more work for you.

   • As a last example, your figures might be meddled with: you might be forced to convert them into obsolete file formats (because apparently, EPS is still the best format available), or, more appallingly, vector graphics might be converted to raster images (judging from the experiences of my friends and myself, this is unfortunately relatively common!). This might sound like a tiny problem again, but it decreases readability and accessibility for some readers, and, more to the point of this post, it is somewhat unnecessary meddling.

   Let me re-iterate my main point: I do not object to changing my paper, I merely object to meddlesome changes that are just generating useless work. For example, there is no need whatsoever to typeset your equations again—this is quite literally the definition of negative work.

5. If you survived this ordeal intact, you now must pay. To be fair, not all journals charge you for normal articles, but most of them charge you for open access publishing. In other words: if I want my research, which is generously funded⁵ by the Swiss taxpayers, to be available to those selfsame taxpayers, I have to pay. The amounts vary a little bit, of course, but we are talking upwards of a few hundred USD at least. Luckily, this is not a problem for my research group; my postdoctoral adviser, Prof. Karsten Borgwardt ensures that sufficient funds for open access publications are available.

   Interestingly, sometimes the cost is fielded by a conference; this happens when the conference has a contract that ensures that its publications will be available as special issue of some journal. This might seem nice because it shifts the costs away from authors, but it is also somewhat non-transparent; conference costs being high already, I find it strange that some of the money goes into the pockets of another party.

   After all this negativity, it is time for a positive example: NeurIPS, one of the flagship machine learning conferences, is partnered with a publisher and makes all papers available for free online. I gladly pay the conference fee for this!

A better model

How can this process be improved? I have a few suggestions:

1. Transparency: publishers should make it clear where the funds are going. Are we increasing shareholder value by working for free? How are profits split and used?

2. Giving back: it is generally understood that everyone needs to eat and no one should have to work for free. Why is then that this completely different in publishing? Almost all the profits are essentially generated because editors and reviewers work for free. I know that being remunerated for your reviewing work might raise some questions about impartiality etc., so I think paying people to write reviews might be somewhat problematic.

   However, closely related to my point about transparency, publishers could be more upfront about how they user their funds and give back to the community. For example, publishers could sponsor students so that they can visit a conference for free, or publishers could sponsor the conferences themselves.

   If you, as a publisher, engage the community and give back a little, the community will be all the more happy to work with you. We need you, but you also need us. Without the scientists, you cannot be a scientific publisher.

3. Commitment to excellence: publishers should commit to the highest quality and the highest standards. Employ people that are capable of working with the scientists, not for the scientists. Train your employees to be experts in typography, typesetting, and pair them with domain experts so that they do not create more work for the authors by inadvertently destroying equations, figures, and so on.

   This goal is not necessarily orthogonal to maximising your profits, by the way: if you lower your standards, your reputation as a publisher will suffer, meaning that scientists in the long run (!) might not be willing to publish with you any more. If you commit to excellence, by contrast, we will flock to you.

   I know that working with a publisher that cares about the end product as much as I do is a heavenly match! So we should endeavour to make such experiences the norm, not the exception.

If publishers work on these three suggestions, they will start to really provide added value to the publishing process by working with the authors in order to find the best way of showcasing good research.

Note that despite my misgivings about some things in the publishing process, I am confident and optimistic that we can build a better way together. The machine learning community, for example, already created their own journal to provide a high-quality, low-friction publishing experience. Let us strive to improve this for other domains as well.

Until next time, good luck with your proofs!